SUMMARY/ABSTRACT This project will fill a fundamental knowledge gap in influenza epidemiology, which is the lack of a quantified relationship between viral antigenic drift and human susceptibility to influenza. We will then apply this new knowledge to improve the accuracy and timeliness of influenza forecasts. Antigenic drift refers to gradual changes in the surface proteins of influenza viruses, which allow new virus strains to escape acquired immunity and to re-infect individuals who were previously infected with influenza. Antigenic cartography can quantify the magnitude of antigenic drift (i.e. the antigenic distance) between influenza virus strains. To date, however, the relationship between antigenic distance and susceptibility to infection has not been quantified for human influenza. We will use a mechanistic model of influenza transmission and immunity to estimate the association between increasing antigenic distance and increasing susceptibility to infection with influenza. For this, we will take advantage of a unique data resource: active influenza surveillance conducted since the 2011/12 influenza season through the US Influenza Vaccine Effectiveness Network. These data include population-based estimates of the incidence of influenza, stratified by virus subtype/lineage and with antigenic and genetic characterization of circulating influenza viruses, in three geographically distinct US states. The data also include influenza vaccine coverage for the target populations. We will apply our mechanistic influenza model to these data and quantify the drift/susceptibility association. We will then apply these findings to improve forecasting of seasonal influenza epidemics. Two different approaches are currently taken to influenza forecasting. Short-term forecasts use near-real-time surveillance data to predict the timing and intensity of the peak in influenza cases, with lead times of a few weeks. Long-term forecasts use data on the relative prevalence of different influenza strains to predict which strains will dominate the upcoming season. At present neither short- nor long-term forecasting methods make effective use of data on pre-existing immunity to influenza due to vaccination or prior circulation of influenza strains. Having quantified the drift/susceptibility association, we will test the forecasting abilities of our influenza model. We hypothesize that including data on prior circulation of influenza and on vaccine coverage will allow us to forecast the intensity and subtype/lineage distribution of upcoming influenza epidemics with lead times of 9+ months. The proposed research will benefit human health by 1) improving our understanding of the interplay between human immunity and virus antigenic drift and 2) improving the accuracy and timeliness of influenza forecasts, allowing more time for the allocation of resources for influenza prevention and treatment.